Combustion and Flame 177 (2017) 37–48

Contents lists available at ScienceDirect

Combustion and Flame

journal homepage: www.elsevier.com/locate/combustflame

Dissociative influence of H 2

O vapour/spray on lean blowoff and NO x

reduction for heavily carbonaceous syngas swirling flames

D.G. Pugh

∗, P.J. Bowen , R. Marsh , A.P. Crayford, J. Runyon, S. Morris, A. Valera-Medina, A. Giles

Cardiff School of Engineering, Cardiff University, Wales, UK

a r t i c l e i n f o

Article history:

Received 7 June 2016

Revised 21 July 2016

Accepted 9 November 2016

Keywords:

Turbulent lean premixed flames

Humidified syngas combustion

Chemical kinetics

OH-PLIF

Emissions

Converter gas

a b s t r a c t

Recent studies have described and evidenced the enhancement of fundamental combustion parameters

such as laminar flame speed due to the catalytic influence of H 2 O with heavily carbonaceous syngas mix-

tures. In this study, the potential benefits of these subtle changes in water loading and hence reaction

pathways are explored in terms of delayed lean blowoff, and primary emission reduction in a premixed

turbulent swirling flame (Ø = 0.6–0.8), scaled for practical relevance. Chemical kinetic models initially

confirm that H 2 O has a substantial impact on the employed fuel behaviour; increasing flame speed by up

to 60% across an experimental range representative of fluctuation in atmospheric humidity ( ∼1.8 mol%).

OH

∗ chemiluminescence and OH planar laser induced fluorescence (PLIF) were employed to analyse the

changes in heat release structure resulting from the experimental addition of H 2 O vapour to the com-

bustor. Equivalent concentrations of liquid H 2 O were introduced into the central recirculation zone of the

premixed flame as an atomised spray, to investigate the influence of phase changes on the catalytic effect.

Near the lean stability limit, H 2 O addition compresses heat release to shorten the elongated flame struc-

ture. Whereas with a stable and well-defined flame structure, the addition triggers a change in axial heat

release location, causing the flame front to retract upstream toward the burner outlet. Higher quantities

of two-phase flow were combined to explore the possibility of employing the spray as a stabilising mech-

anism, effectively dam pening the observed influence of humidity. The chemical enhancement induced by

the controlled supply was shown to reduce the lean blowoff stability limit, enabling an increase in ad-

ditional air flow of almost 10%. However, the catalytic effect of H 2 O diminishes with excessive supply

and thermal quenching prevails. There is a compound benefit of NO x reduction from the use of H 2 O as a

flame stabiliser with the practically-relevant syngas: First NO x production decreases due to thermal effect

of H 2 O addition, with potential for further reduction from the change in lean stability limit; leanest ex-

perimental concentrations reduced by up to a factor of four with two-phase flow at the highest rates of

supply. Hence, the catalytic effect of H 2 O on reaction pathways and reaction rate predicted and observed

in the laminar environment, is shown to translate into practical benefits in the challenging environment

of turbulent, swirl-stabilised flames.

© 2016 The Authors. Published by Elsevier Inc. on behalf of The Combustion Institute.

This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ).

1

8

T

t

v

[

f

h

d

a

c

r

fl

r

h

0

(

. Introduction

Heavily carbonaceous converter gas typically comprises 50–

0% CO, 10–18% CO 2 , 1–3% H 2 in a balance of N 2 (mol%) [1,2] .

he influence of water content on combustion performance for

his increasingly utilised fuel is highly relevant [2–9] , as it pro-

ides a potential dissociative catalytic influence on CO oxidation

3,4] across a limited water loading range. Direct formation of CO 2

∗ Corresponding author.

E-mail address: pughdg@Cardiff.ac.uk (D.G. Pugh).

i

w

l

ttp://dx.doi.org/10.1016/j.combustflame.2016.11.010

010-2180/© 2016 The Authors. Published by Elsevier Inc. on behalf of The Combustion In

http://creativecommons.org/licenses/by/4.0/ ).

rom CO is slow due to high activation energy, and the presence of

ydrogen facilitates chain branching OH formation, changing the

ominant path for oxidation [4] . This contrasts with a lowering of

diabatic flame temperature. Water addition therefore induces a

ompetition between chemical and thermal effects over a limited

ange. This provides a non-monotonic influence on premixed

ame propagation, increasing laminar flame speed and reaction

ate at lower concentrations, before eventually diluting and slow-

ng the reaction as loading increases. Recent studies undertaken

ith a converter gas mixture demonstrated the potential for

arge increases in laminar flame speed – up to 70% – with the

stitute. This is an open access article under the CC BY license

38 D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48

r

(

o

A

o

p

r

(

c

(

t

c

n

o

f

2

t

t

a

A

p

c

e

2

d

a

a

p

u

m

i

w

w

p

v

i

fi

c

m

t

e

s

t

i

a

t

i

fi

v

c

a

fl

s

(

i

e

s

i

d

o

t

O

addition of water vapour, as concentrations varied by amounts

representative of change in atmospheric humidity [4] . This level of

fluctuation would be expected to provide a significant impact on

the behaviour of converter gas in practical application.

1.1. Aim of this work

With previous emphasis on the fundamental laminar premixed

behaviour of the applied fuel mixture [3,4] , the aim of the current

investigation is to appraise to what extent these potentially posi-

tive features extend to the practical environment of turbulent, pre-

mixed swirling flames supplied with a similar representative fuel

composition. Clearly whilst the laminar flame characteristics pro-

vide an indication of fuel behaviour in turbulent swirling flames,

these also introduce further complexity, for example the influence

on flow dynamics and flame stretch [10,11] . Nevertheless, in a qual-

itative sense, the laminar studies indicate that practical benefits

such as extended flame blowoff limit, and improved NO x emission

performance for swirl burners are likely [12–15] . Variation in these

performance characteristics are quantified within this study.

Water vapour concentration was varied initially across a range

equivalent to change in atmospheric humidity, as a fraction of

the total reactant flow (up to ∼1.8 mol% of total reactant com-

position). This allowed potential fluctuations in flame behaviour

to be studied, within a range of practical relevance. This change

in behaviour was analysed using advanced optical diagnostic

techniques; namely OH

∗ chemiluminescence, and OH planar laser

induced fluorescence (PLIF) [16–18] enabling evaluation of any

variation in local heat release, in addition to the formation of

intermediate OH previously shown to accelerate the reaction [3,4] .

The net thermal power of the flame from the generic swirl burner

employed was maintained at 100 kW throughout.

The experimental setup was designed so water could be intro-

duced into the system either as a vapour or an atomised liquid

spray – delivered into the central recirculation zone of the swirling

flow – or both. Potential differences in the catalytic behaviour

resulting from the H 2 O phase change could then be studied for its

direct influence on flame stability. Finally, two-phase parametric

studies were conducted to examine the effectiveness of employing

a central recirculation spray to reduce the vapour-induced en-

hancements in dry flame behaviour. In addition to flame stability,

the production of gaseous emissions was also quantified, with

attention given to NO x formation.

2. Experimental facilities

2.1. Burner assembly

Experiments were undertaken using a premixed generic swirl

burner, housed within a high pressure optical chamber (HPOC)

at Cardiff University’s Gas Turbine Research Centre (GTRC). The

burner and casing assembly are shown as a multi-component

schematic in Fig. 1.

The burner assembly has been modified from previous work

[16] , with the addition of a centreline liquid injection lance

( Fig. 1 (a)); water is supplied from a pressurised 20 L accumulator

in the range of 0.7–2.2 MPa, with delivery measured using an

inline Coriolis mass-flow meter (Emerson CMF010M ± 0.1%). Reac-

tants enter the burner inlet plenum ( Fig. 1 (b)) via fixed piping or

flexible metallic hose, with mass flows again quantified using Cori-

olis meters (Emerson CMF025M ± 0.35%). When steam is required,

liquid flow is regulated using a Bronkhorst mini CORI-FLOW con-

troller ( ± 0.2%), before changing phase in a 25 m heated supply

line. The burner assembly is housed within a pressure casing

( Fig. 1 (c)) rated to 1.6 MPa at 900 K. After entering the plenum,

reactants travel through the premix chamber ( Fig. 1 (d)) to a single

adial-tangential swirler ( Fig. 1 (e)) and out the burner exit nozzle

20 mm radius). The outlet has a geometric swirl number (ratio

f tangential to axial momentum) equivalent to S g = 1.04 [19,20] .

Delevan 0.4 hollow cone (W) 60 ° spray nozzle was installed

n the liquid injection lance to provide atomised water droplets,

ositioned to spray into the CRZ of the outlet flow field, and

eleased in the same plane as the 40 mm diameter burner outlet

Fig. 1 (f) – spray is characterised further in Section 3 ). Optical ac-

ess is afforded via diametrically opposed quartz viewing windows

Fig. 1 (g)), facilitating the chemiluminescent and PLIF diagnostic

echniques. The burner is operated with cylindrical quartz exhaust

onfinement tube, with an expansion ratio of 3.5 from the burner

ozzle exit ( Fig. 1 (h)). The exhaust is then sampled downstream

f the confinement in a temperature conditioned probe, detailed

urther in Section 2.4 . Several k and n-type thermocouples ( ±.2 K) were installed to monitor key rig temperatures, influential

o the experimental output. The first thermocouple is housed in

he inlet plenum (T1) feeding into the premix swirl chamber,

nd monitors the temperature of reactants entering the system.

second thermocouple mounted flush with the burner face (T2)

rovides a reading for the recirculated and radiated heat from

ombustion. Finally, a third measurement is made at the burner

xit (T3) giving exhaust temperature at the quartz tube outlet.

.2. OH

∗ chemiluminescence measurements

The chemiluminescence of an excited hydroxyl (OH) interme-

iate radical (wavelength of ∼310 nm) can be used to provide

non-intrusive qualitative indication of localised heat release,

nd a generalised marker of flame front location [16–18] . For the

urposes of this work, the relatively weak OH

∗ signal was captured

sing a Dantec Dynamics Hi Sense Mk II CCD camera with a 1.3

egapixel resolution, coupled to a Hamamatsu C9546-03 L image

ntensifier. A specialty 78 mm focal length UV lens (F-stop = f /3.8)

as installed on the aforementioned image intensifier, together

ith a narrow band pass filter (300–330 nm). The camera is

laced at a 90 ° angle to the flow such that the flame can be

iewed through the top quartz window of the HPOC ( Fig. 1 ). The

mage plane is centred on the burner exit, with respective view

elds of ∼100 mm and ∼75 mm in the radial and axial directions,

orresponding to a resolution of ∼13.6 pix/mm. For each experi-

ent, 200 images were captured at a rate of 10 Hz with the system

iming controlled using commercial software and a gate pulse gen-

rator set at 400 μs, triggered by the camera aperture signal. To en-

ure consistency between OH

∗ chemiluminescence measurements,

he gain of the image intensifier was held constant for all exper-

ments. The intensities measured for the 200 image dataset were

veraged, with a condition-normalised false colourmap (correlated

o relative pixel signal intensity) then applied, to help provide an

ndication of the flame front location and localised heat release.

The axial flow field generated by the employed swirl con-

guration has been previously evaluated using particle image

elocimetry [21] , with a planar representation at the nozzle

entreline shown in Fig. 2 (a). The flame structure is typical for

premixed swirler of this design [22] , with outward forward

ow (orange in Fig. 2 (a)) generating a conical flame around a

hear layer of zero axial velocity, and central recirculation zone

CRZ – blue in Fig. 2 (a)). The observed OH

∗ chemiluminescence

s taken line-of-sight, meaning the captured image includes light

mitted both in front, and behind the focal plane of the imaging

ystem. Thus, in the case of the swirl-stabilised flames evaluated

n this work, the resulting images show intensity from the three-

imensional, conical shape of the flame ( Fig. 2 (b)). A modified

pen access Abel inversion algorithm [23] was used to transform

he 3-D images captured into 2-D planar representations of the

H

∗ chemiluminescence intensity distribution, with the underlying

D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48 39

Fig. 1. Multi-component schematic of the burner assembly and optical casing (components are described in the text).

b ca

Fig. 2. Simplified planar representation of the axial flow field at the nozzle centreline (a), with a comparison between full OH

∗ flame chemiluminescence (b), and the

equivalent Abel transformed image (c).

a

i

a

e

f

b

D

c

2

i

f

Q

o

o

t

fl

i

n

l

o

t

c

t

a

1

t

m

s

i

t

t

g

b

e

t

c

a

l

P

c

c

ssumption that the swirl-stabilised flame is axisymmetric about

ts centreline ( Fig. 2 (c)) [22–25] . Both raw chemiluminescence,

nd Abel deconvoluted images are presented for the undertaken

xperiments, with an example of the symmetric flame shown

or comparison in Fig. 2 (figure is oriented so flow is from the

ottom up, with intensities normalised to the image maximum).

etailed information regarding the chemiluminescence image

apture system and processing technique can be found in [25] .

.3. Planar laser induced fluorescence (PLIF) measurements

In addition to OH

∗ chemiluminescence, qualitative planar laser

nduced fluorescence (PLIF) measurements were also obtained

or each experimental condition. The PLIF system comprises a

uantel TDL-90 dye laser pumped by the 532 nm output beam

f a frequency-doubled Spectra Physics GCR 170-10 Nd:YAG laser

perating at 10 Hz. The output beam of the TDL-90 was tuned

o 283 nm to excite the (1,0) band of the OH radical, while

uorescence was captured at 315 nm through the use of the same

ntensified CCD camera system as listed for the OH

∗ chemilumi-

escence measurements above. The output beam from the dye

aser was then directed through a set of telescopic sheet-forming

ptics to provide a laser sheet 25 mm in width. This entered

he test rig through a side window of the HPOC while the CCD

amera was focussed on the burner nozzle centreline, through the

op window of the HPOC, perpendicular to both the laser sheet

nd fluid flow. This system produces pulse energies of roughly

6 mJ/pulse at 283 nm, which is critical for the test rig setup given

he size and shape of the HPOC windows, and the quartz confine-

ent tube. Given the short lifetime of the OH radical fluorescence

ignal and influence of quenching effects, the gate timing of the

mage intensifier was set to 1.5 μs, triggered by the Q-switch of

he Nd:YAG laser to ensure that signal capture was appropriately

imed with the laser excitation light pulse. The image intensifier

ain was held constant at all test conditions to ensure compara-

ility, while reducing signal to noise ratio as much as possible. For

ach test condition, 500 images were taken at a rate of 10 Hz, with

he same approximate view fields and pixel resolution as the OH

∗

hemiluminescence images. The obtained image intensities were

gain temporally averaged, and corrected for background reflected

ight for presentation. The uncertainties in obtaining quantitative

LIF measurements are significant, with correction required for

hanges in collisional quenching and temperature [26] for each

ondition of the employed experimental specification. Qualitative

40 D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48

a

a

r

p

e

a

t

a

P

w

t

S

e

‘

t

p

m

e

o

b

e

t

I

t

r

e

t

n

c

i

v

r

4

4

[

c

c

m

D

[

c

o

b

t

a

a

c

i

a

t

t

t

a

c

s

F

fl

t

h

results were therefore analysed for relative change in normalised

OH concentration profiles as markers for flame location and

stability, similar to other equivalent works [22,27] .

2.4. Exhaust emissions measurements

The exhaust stream was sampled using an equal area probe (9

holes) water conditioned to 433 K, and located approximately 10

diameters downstream of the quartz combustor exit, in accordance

with ISO 11042 [28] . The probe, sample line, filter and distribution

manifolds were maintained at 433 K, with a heated pump used

to deliver products to the analysis suite. Total NO x concentra-

tions were quantified using a heated vacuum chemiluminescence

analyser (Signal Instruments 40 0 0VM), calibrated in the range

0–39 ppmV. Concentrations were measured hot and wet to avoid

any water absorption losses, with data normalised to equivalent

dry conditions prior to presentation. A flame ionisation detector

(Signal Instruments 30 0 0HM) was also employed to measure

unburned hydrocarbons at C 3 H 8 equivalent concentrations in the

heated gas stream. Additional sample was directed to a chiller,

used to reduce the molar water concentration below 1% before

downstream CO and O 2 measurements were made using a com-

bination of nondispersive infrared and paramagnetic analysers

(Signal instruments 90 0 0MGA). These were calibrated in the

respective ranges of 0–904 ppmV, and 22.52%vol.

3. Specification of experimental conditions

Fuel fractions were taken from previous work [2,4] , and com-

prised a molar composition of 65% CO, 1% H 2 and a combined

diluent fraction of 34% N 2 . All measurements were made at atmo-

spheric pressure, and 100 kW net thermal power (a constant fuel

mass flow rate of 14.9 g s −1 ) for three initial equivalence ratios (Ø)

in the range 0.6–0.8 (air flows of 30.4–40.2 g s −1 ). This was spec-

ified with consideration given to performance of the liquid spray

nozzle: A maximum flow of 0.52 g s −1 could be achieved from

the accumulator at 2.2 MPa supply pressure, and the experiment

was designed so that water concentrations were varied to levels

representative of change in atmospheric humidity. Allowing for

the equivalence ratio specification, 100 kW thermal power meant

volumetric flow increases of 1.5–1.8% with water addition. With re-

spect to the overall reactant composition; this was equivalent to a

relative humidity of ∼50% at 298 K, or fully saturated at 288 K, and

therefore offered conditions in a range representative of practical

application. Fully vaporised, the addition of water also resulted

in a most significant (realised at Ø = 0.8, when relative fraction

is highest as airflow is reduced) density reduction of 0.65%, and

isobaric heat capacity increase of 0.27% for the overall reactant

mixture. The influence of water addition on flow dynamics was

therefore minimal compared to the change in chemical timescales.

The design characteristics of the nozzle also provided a

limitation on the lowest possible flow sprayed into the flame,

corresponding to 0.32 g s −1 at 0.7 MPa. Commissioning trials were

performed, and supply pressures under this value resulted in poor

atomisation and a limited spray cone definition. Therefore an in-

termediate test point equivalent to 0.42 g s −1 was finally specified

for the liquid spray case. Water droplet SMDs were expected to

vary between approximately 75 and 90 μm [29] for the given noz-

zle and pressure range specification, with ∼17% smaller droplets

produced at the highest pressure. However it should be noted that

data were obtained without a flame, with performance at hotter

conditions unquantified. For the fully vaporised experiments, the

same flow rates were defined for comparison, with the exception

of the 0.42 g s −1 case. Instead the flow was reduced by the equiv-

alent full scale margin to 0.12 g s −1 , to provide an intermediate

measurement with a lower flow rate. Finally both methods of H O

2

ddition were combined to give a maximum of 0.94 g s −1 , and

nalyse the spray influence on an already humidified premixed

eactant flow.

Fuel was supplied from gravimetrically blended cylinder packs,

roduced in the same batch with a 1% full-scale uncertainty of

ach mixture component. The cylinders were constructed from

luminium to minimise the potential for iron pentacarbonyl forma-

ion from the pressurised storage of carbon monoxide. Combustion

ir was supplied from an Atlas Copco GA 45 VSD compressor.

rior to entering the distribution pipework, the compressed air

as dried using a Beko Drypoint DPRA960. The dew point at

he inlet plenum was measured using a Michell Instruments

30 0 0 hygrometer, returning values in the range of −19 to −17 °Cquivalent to ∼1300 to ∼1600 ppmV H 2 O, which gave the baseline

dry’ condition. In addition, combustion air was analysed using

he flame ionisation detector for hydrocarbon contamination as

art of the experimental setup procedure. Concentrations were

inimal, and typically measured in the range 0–1 ppmV C 3 H 8

quivalent. The entire experimental system was warmed for up to

ne hour to reduce some of the effects of local heat loss, evaluated

y monitoring the exhaust temperature, T3 in Fig. 1 . Nominally

quivalent measurements were taken at various stages throughout

he programme with minimal temperature variation noted ∼2%.

n addition, the inlet plenum temperature (T1, Fig. 1 ) was moni-

ored, and typically ∼300 ± 3 K, with an increase of up to ∼20 K

esulting from the introduction of steam to the premixed reactants.

The lean stability limit (lean blowoff, or LBO point) was also

valuated for each experimental condition. This was defined by

he equivalence ratio that resulted in detachment from the outlet

ozzle to give a lifted flame, stabilising downstream in the quartz

onfinement tube with increased air flow. This was obtained by

ncrementally increasing air flow via a digitally controlled needle

alve giving steps of ∼0.2 g s −1 , and visually monitoring the flame

esponse.

. Results and discussion

.1. Modelling reaction kinetics

Chemical kinetic models were obtained using the PREMIX

30] program in the CHEMKIN-PRO [31] laminar flame speed

alculator to provide 1-D simulations of intermediate reaction

hemistry, and help understand and analyse the expected pre-

ixed flame response. The calculation employed the modified

avis et al. [32] reaction mechanism (MDM) developed recently

4] , optimised for use with humidified syngas compositions, and

omprises 14 chemical species and 38 reactions. Unlike the previ-

us work, this study focussed on lean conditions to explore lean

lowoff in particular, and hence combustion was modelled for the

hree specified equivalence ratios (Ø = 0.6, 0.7, and 0.8). Solutions

re based on an adaptive grid of 10 0 0 points, with mixture-

veraged transport properties and trace series approximation. For

onverter gas, water has been shown to provide two competing

nfluences on premixed flame propagation; first it acts as a diluent

nd the reaction is slowed by a reduction in adiabatic flame

emperature [4] . The second effect results from changes in flame

hermochemistry, and an increase in overall reaction rate due to

he dissociative production of chain carrying species (such as OH),

nd enhanced heat release rate. The catalysing effect of these radi-

als on CO consumption processes have been shown to reduce the

low reaction (with high activation energy): CO + O 2 → CO 2 + O [2,4] .

igures 3 and 4 show the respective changes in predicted adiabatic

ame temperature and maximum heat release rate (HRR) for

he employed fuel composition. Note that for consistency results

ave been plotted against overall mixture H O fraction, with the

2

D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48 41

Fig. 3. Change in predicted adiabatic flame temperature against mixture H 2 O frac-

tion for the employed converter gas composition at Ø = 0.6, 0.7, and 0.8 with ex-

perimental markers.

Fig. 4. Change in predicted maximum HRR against mixture H 2 O fraction for the

employed converter gas composition at Ø = 0.6, 0.7, and 0.8 with experimental

markers.

e

a

m

e

r

i

t

f

Fig. 5. Change in modelled laminar flame speed against mixture H 2 O fraction for

the converter gas composition employed at Ø = 0.6, 0.7, and 0.8, with experimental

markers.

a

s

f

T

r

Ø

r

c

Ø

l

m

fl

p

f

b

a

u

e

t

f

o

r

p

d

t

r

i

w

4

fl

s

fl

i

xperimental mass water loadings specified in Section 3 identified

s markers for each configuration.

Adiabatic flame temperature is shown to reduce by approxi-

ately ∼20 K with water loading for each of the three modelled

quivalence ratios, and the individual vapour and spray configu-

ations. In comparison, maximum heat release rate significantly

ncreases across the experimental range, more than doubling for

he Ø = 0.8 experimental condition, and appearing to peak at H 2 O

ractions of 0.05–0.06. The overall response is for a faster reaction,

s demonstrated with the changes in laminar flame speed pre-

ented in Fig. 5 where again, data have been plotted against H 2 O

raction, with the experimental data points identified as markers.

he increase in flame speed is greater for the richer equivalence

atios; with the highest water loading giving a rise of ∼60% for

= 0.8, compared to ∼53% for Ø = 0.6. This results from the

elative difference in molar H 2 O fraction of the overall reactant

omposition, i.e. ∼1.8 mol% for Ø = 0.8, compared to ∼1.5 mol% for

= 0.6. This emphasises the impact that subtle changes in water

oading can have with this fuel blend, and suggests that richer

ixtures will provide a more significant change in the observed

ame response. The increase in flame speed is non-monotonic,

eaking at a H 2 O fraction of ∼0.05. As water loading is increased

urther, enhancement in HRR diminishes, temperature suppression

ecomes dominant with a ∼75 K reduction in AFT, and the re-

ction slows [4] . This has been further quantified experimentally

sing laminar flames with similar mixtures [3,4] .

OH was used as a marker for the optical diagnostic techniques

mployed in the experimental study, and it was therefore impor-

ant to analyse any potential changes in production that result

rom H 2 O addition. Figure 6 demonstrates the influence of water

n the 1-D spatial concentration profiles for two equivalence

atios; Ø = 0.6 and 0.8. Water addition is shown to enhance the

roduction of OH with the given fuel blend through intermediate

issociation, resulting in peak wet concentrations that are almost

riple the equivalent dry cases. In addition to catalysing the overall

eaction, this increased production of chain carrying radicals also

ndicated that a broad range of concentrations, or signal intensities

as to be expected from the analytical diagnostics.

.2. Vapour addition to the 100 kW swirl burner

The supply of water vapour was increased at the specified

owrates, with the OH

∗ chemiluminescence response for Ø = 0.6

hown in Fig. 7 - the global signal for projection of half the 3-D

ame is shown (a), together with the equivalent Abel inverted

mage (b), with flow from the bottom up. To improve clarity in

42 D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48

Fig. 6. Modelled 1-D OH spatial concentration profiles for Ø = 0.6 and 0.8, with 0

and 0.52 g s −1 H 2 O.

a

b

Fig. 7. OH

∗ chemiluminescence signals for projection of the 3-D half flame (a) and

equivalent Abel transformed images (b), for Ø = 0.6 with increasing levels of vapour

addition.

a

b

Fig. 8. OH

∗ chemiluminescence signals for projection of the 3-D half flame (a) and

equivalent Abel transformed images (b), for Ø = 0.8 with increasing levels of vapour

addition.

a

fl

a

w

b

h

s

o

w

fl

Ø

c

b

P

t

d

0

i

s

a

p

b

a

m

c

fl

t

s

e

c

p

r

C

presentation, intensities are normalised to the maximum intensity

in the average image for each condition, with equivalent propor-

tions maintained for all images at the specified scale. The flame

was close to the LBO stability limit at the driest condition, with

the poorly defined structure appearing elongated. The higher OH

∗

intensities present in the downstream flow are expected, resulting

from the overall slower heat release and reaction rates. There also

ppears to be higher relative recirculation outside the core forward

ow. As the vapour fraction increases, chemical timescales reduce

nd the flame compresses into a more defined conical structure,

ith outward flow strengthening and the tip of the inner CRZ

ecoming more prominent. Chemical kinetics suggest that relative

eat release increases with reaction rate – evident outside the

hear layer of the premixed swirling flow with marginal thickening

f this region observed in the Abel transformed images.

The equivalent images are shown for Ø = 0.8 in Fig. 8 , again

ith a comparison made between the temporally averaged 3-D

ame (a), and equivalent Abel transform (b). Compared to the

= 0.6 case, heat release in the richer flame does not appear to

ompact in the same way with the structure already well-defined,

ut instead transitions further upstream as reaction rate increases.

lanar OH profiles were also obtained using the PLIF system with

he 25 mm-wide laser sheet, fixed at an axial position 13 mm

ownstream of the premixed nozzle. These are shown for Ø =.6, and 0.8 at each averaged vapour condition in Fig. 9 (note –

nstantaneous images were processed in the same way, but were

hown to endorse the explanation derived from the averaged set

nd hence are not included). The images show the measured

lanar OH fluorescence in a horizontal plane positioned at the

urner centreline. The unburned reactants from the outward flow

re identified by the regions of low signal intensity, with OH

easured in the inner and outer recirculation zones. Higher OH

oncentrations are ostensibly recirculated outside of the forward

ow for the driest conditions, with a more uniform intensity dis-

ribution across the CRZ. As water vapour concentration increases,

ignal intensities increase upstream closer to the burner exit, as

xpected from the increase in reactivity and OH

∗ chemilumines-

ence results. However, there is also a change in the location of

eak relative OH concentration; inasmuch that as vapour flow

ises, higher intensities are located inside the shear layer of the

RZ. The 1-D spatial concentration profiles ( Fig. 6 ) suggest a

D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48 43

Fig. 9. OH PLIF intensities for Ø = 0.6 and 0.8 with increasing levels of water vapour addition.

a

b

Fig. 10. OH

∗ chemiluminescence signals for projection of the 3-D half flame (a) and

equivalent Abel transformed images (b), for Ø= 0.6 with increasing levels of water

spray.

s

c

p

a

b

4

o

r

fl

Ø

fl

a

s

C

a

b

Fig. 11. OH

∗ chemiluminescence signals for projection of the 3-D half flame (a) and

equivalent Abel transformed images (b), for Ø = 0.8 with increasing levels of vapour

spray.

c

a

s

t

o

v

F

A

v

s

m

r

i

a

p

harp increase in OH for the wettest conditions followed by a fall,

ompared to the drier cases where the OH signal reaches a relative

lateau. It therefore follows that higher relative concentrations

re found closer to the reacting flow, near the axial shear layer

oundary and outer recirculation zones.

.3. Liquid spray addition to the 100 kW swirl burner

The spray angle of the liquid nozzle (60 °) was chosen to deliver

ffset flow into the CRZ, recirculated into the forward flow of

eactants. OH

∗ chemiluminescence data were captured for each

ow condition, shown with corresponding Abel deconvolution for

= 0.6 in Fig. 10 . Again, the images show half the symmetrical

ame, and have been resized for presentation (fixed proportions

re maintained for all images presented, equivalent to those

hown in Figs. 7 and 8 ). As the droplets are delivered into the

RZ, the flame experiences enhancement due to the effects of

atalytic chemical dissociation, contrasted against heat loss and

change in local strain and turbulence intensity resulting from

pray dynamics and droplet transportation [33] . The net effect is

o compact and narrow the flame structure, shortening the area

f localised heat release by a greater amount than the equivalent

apour experiments.

Similar influences are evident for the Ø = 0.8 case shown in

ig. 11 , as the flame retracts upstream and spray flowrate increases.

gain, overall flame size reduces compared with the equivalent

apour cases shown in Fig. 8 . The Abel inverted images suggest a

hortening of the flame structure, as peak heat release intensities

ove from the boundary confinement, chemical timescales are

educed locally and the CRZ contracts, trends that are also evident

n the OH PLIF results: Fig. 12 (a) gives a comparison between the

verage Ø = 0.8 spray and vapour experiments with the laser

lane fixed in an equivalent position; ∼13 mm downstream of the

44 D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48

a b

Fig. 12. (a) Comparison between centreline OH PLIF planes positioned 13 mm downstream from the burner nozzle, for equivalent vapour and spray experiments at Ø = 0.8.

(b) Influence of water spray addition on centreline OH PLIF planes at different downstream axial locations at Ø = 0.6, note the difference in y-axis scales.

Fig. 13. Change in measured exhaust (T3) and burner face temperatures (T2) re-

sulting from water addition.

fl

c

4

i

fl

p

t

w

i

r

i

i

a

w

burner exit nozzle. OH fluorescence from the spray experiments

indicate a marginal narrowing of the CRZ with an overall width

reduction of 12–16% (calculated from the geometry of the scaled

mages), and larger areas of recirculated OH present outside the

forward flow. In addition, higher OH intensities are observed again

on the outer boundary of the CRZ, near the axial shear layer. This

is present in greater detail in Fig. 12 (b), where data have been

collected at different axial locations in the Ø = 0.6 flame.

The 25 mm-wide laser sheet was repositioned at downstream

distances of 5 mm and 35 mm, on the centreline from the burner

exit nozzle. Further data was collected for the driest experiments,

with the introduction of the spray again causing a relative increase

in the average measured upstream OH intensities, as reaction rate

increases and the flame changes position. This is also apparent in

the data collected 35 mm downstream, with the retracted flame

wider for the wetter test. The dry case appears to have a more con-

sistent OH intensity distribution across the tangential flame profile,

with peaks forming in the outer layer of the CRZ at the 0.32 g s −1

condition. Again, this is likely to result from peaks in OH produc-

tion profiles observed in Fig. 6 , relative to the localised outward

reacting flow. Introduction of the spray is expected to cause some

inhomogeneity in the reaction zone as H 2 O is entrained into the

recirculated flow, and may also result in localised OH production.

In addition to inhomogeneity and change in phase, there are

other subtle differences between the two experimental config-

urations which influence the flame. The increase in bulk exit

velocities (1.8% worst-case) and flow dynamics resulting from the

introduction of vapour apply to the entire outward flow field.

Whereas the premixed gaseous exit velocities remain constant

for the spray experiments, with the additional flow, and change

in turbulence intensity, introduced locally to the CRZ from water

dispersion and transportation. In addition, the introduction of a

liquid spray further reduces flame temperature by an additional

∼20 K for the highest flowrates, resulting from the latent heat

required for phase change. This is shown in Fig. 13 , where compar-

isons of the measured combustor exhaust temperature (T3) have

been plotted – averaged for all equivalence ratios for simplicity in

presentation – against experimental water mass loading. Burner

face temperatures (T2) measured during the vapour experiments

have also been included. An expected temperature rise is noted as

vapour concentration increases, and the flame retracts upstream

toward the outlet nozzle and burner face. This demonstrates the

significant (20 0–30 0 K) fluctuation in operational temperatures

that may be experienced with the utilisation of this fuel blend

from subtle changes in atmospheric humidity. The combined effect

of this change in temperature with reaction rate also creates

the potential for further change in flame topology or premixed

ashback, which could possibly occur with a different burner

onfiguration, or a change in outlet velocities [34,35] .

.4. Two-phase vapour spray addition

The combined influence of two-phase vapour spray flow was

nvestigated to further analyse the mechanisms through which the

ame structure may be influenced by water addition, so enhancing

remixed operational stability. Figure 14 shows a comparison be-

ween the OH

∗ chemiluminescence intensities obtained at Ø = 0.7

ith equivalent flow conditions for the vapour and spray exper-

ments, in addition to both combined at the highest vapour flow

ate. Again, the global signal for projection of half the 3-D flame

s shown (a) together with equivalent Abel deconvolution (b), and

mage proportions maintained from results presented previously.

The heat release trends evident for each of the distinct vapour

nd spray experiments correspond to those presented earlier,

ith the overall structure compacting and the flame retracting

D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48 45

a

b

Fig. 14. OH

∗ chemiluminescence signals for projection of the 3-D half flame (a) and equivalent Abel transformed images (b), for Ø = 0.7 with increasing levels of two-phase

water addition.

u

n

s

c

t

r

o

t

fl

r

o

t

c

l

e

fl

w

t

S

o

r

i

c

e

f

r

l

S

fi

r

u

Fig. 15. Change in LBO equivalence ratio for all experimental configurations.

o

t

l

pstream in the flow. Spray addition is again shown to produce

arrowing of the heat release profile, drawing in the CRZ and

hortening the flame. Operation was shown to be stable at all

onditions of two-phase flow considered, with a significant charac-

eristic observed; initially introduction of the spray causes the heat

elease structure to compact from the equivalent 0.52 g s −1 vapour

nly case. This may be expected given the changes observed from

he spray experiments. However, as liquid flow is increased the

ame structure is shown to expand and elongate, giving closer

esemblance to the vapour only results. This tendency was also

bserved for the Ø = 0.6 and 0.8 experiments, and suggests that

he narrowing of the shortened flame results more from the local

atalytic influence of H 2 O introduction to the CRZ, than the turbu-

ence induced from droplet introduction in the spray. The chemical

nhancement resulting from dissociation has peaked at the highest

ow conditions, and is in competition with the thermal effect of

ater acting more as a traditional diluent, increasing chemical

imescales and slowing the reaction [4] , as quantified in Figs. 3 –6

ection 4.1 . Therefore at higher flow conditions the local influence

f H 2 O on the CRZ is lessened, and the heat release structure

everts back to the vapour form. An equivalent peak was observed

n the burner face temperature (T2), with values falling as the

ombined flow began to elongate the flame.

The catalytic enhancement in flame speed also provided an

xtension of the LBO stability limit, with detachment of the flame

rom the anchor. As the chemical structure compacted and overall

eaction rate increased, air flows could be driven higher to give

eaner equivalence ratios. Using the methodology outlined in

ection 3 , the LBO stability limit was determined for each con-

guration and flow condition with the corresponding equivalence

atios plotted in Fig. 15 , where error bars represent the combined

ncertainty in flow metering and fuel delivery

Water addition is shown to expand the lean stability envelope

f the given fuel blend. Stable air flows increase by almost 10% for

he addition of 0.52 g s −1 water, with vapour delivery providing

eaner limits compared to the equivalent spray conditions. This

46 D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48

a

b

Fig. 16. OH

∗ chemiluminescence signals for projection of the 3-D half flame (a)

and equivalent Abel transformed images (b) for LBO conditions (V-vapour, S-spray,

T-two-phase).

Fig. 17. Change in normalised (dry,15% O 2 ) NO x concentration with water loading

for Ø = 0.6–0.8.

s

(

e

d

θ

(

p

θ

t

c

c

l

s

a

c

t

a

d

[

r

H

r

e

n

a

l

s

t

r

v

i

(

is attributed to the additional heat loss experienced from the

phase change necessary with the liquid supply, and potential

inhomogeneity in the reaction zone. The two-phase water case

demonstrates a shift in the trend, with an increase in the lean

stability limit equivalence ratio. Kinetic results demonstrated

the catalytic enhancement saturating at these flowrates, with

flame temperature and heat release rate reduced, and chemical

timescales extended. It could therefore be projected that a further

increase in water supply would limit the stability envelope further,

eventually reducing it below the driest case. A permanent water

spray therefore has the potential to reduce some of the operational

fluctuations that may result from change in atmospheric humidity,

but flow rates have to be carefully specified. OH

∗ chemilumines-

cence was employed to characterise heat release distribution at

the wettest conditions at the stable lean limit, with results shown

in Fig. 16 . There is an increase in heat release recirculated outside

of the forward flow for all conditions, as the flame extends into

the outer recirculation zones. The vapour and two-phase experi-

ments show the flame elongated, and the structure more apparent

than the dry Ø = 0.6 case ( Figs. 7 and 10 ). However, the local

catalytic influence of H 2 O in the CRZ gives the spray condition the

narrow structure previously observed, and the corresponding Abel

deconvolution shows a larger zone of heat release. Detachment of

the spray flame can therefore be attributed to localised heat loss

near the anchor at burner outlet, as opposed to a reduction of

global chemical reactivity seen with the vapour case, and hence

why the stability limit is comparatively reduced.

4.5. Emission measurements

Emission measurements were made using the analytical setup

outlined in Section 2.4 . Total NO x concentrations were measured

hot and wet, and data were therefore normalised in accordance

with BS EN 11,042 [28] prior to presentation: First the mea-

urement ( θmeas ) was normalised to an equivalent dry reading

θdry ) using the corresponding exhausted water fraction ( X H 2 O ) for

ach experimental condition, obtained from the chemical models

etailed in Section 3 , shown in Eq. (1) .

dry = θmeas ·(

1

1 − X H2O

)(1)

The dry component was then normalised equivalent to 15% O 2

θdry,15%O2 ) dilution using the measured O 2 concentrations ( θO2 ),

resented in Eq. (2) .

dry , 15 %O 2 = θdry ·(

20 . 9 − 15

20 . 9 − θO2

)(2)

CO was analysed dry and therefore only required normalising

o the equivalent 15% O 2 condition. Normalised total NO x con-

entrations were obtained for the vapour and spray experimental

onfigurations, and are shown for Ø = 0.6–0.8 in Fig. 17 trend

ines have been superimposed for clarity. The error bars repre-

ent total uncertainty in the measurement system, and comprise

nalyser specifications, linearisation, and accuracy in span gas

ertification. H 2 O introduction is shown to approximately halve

he normalised NO x concentrations for each equivalence ratio

cross the 0–0.52 g s −1 range. This reduction results from a

rop thermal NO x production with adiabatic flame temperature

8,13–15] , and also provides the offset between each equivalence

atio. Work undertaken by Zhao et al . [13] demonstrated that

2 O can lead to a kinetic increase in NO x production from the

ise in OH concentration at constant temperature, however this

ffect is eclipsed by thermal influences and HCN formation. It is

oteworthy that for the spray experiments, NO x concentrations

re consistently higher than the equivalent vapour case. Even with

ower global temperatures resulting from H 2 O phase change, the

pray experiments feed locally into the CRZ and therefore poten-

ially allow for the formation of localised hotter, drier combustion

egions. Similar results were obtained by Furuhata et al. [36] with

ariation in NO x concentrations resulting from changes in steam

njection location, and localised hotter regions in the flame.

With water addition shown to facilitate leaner operation

Section 4.3 ), the potential for further NO x reduction was

D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48 47

Fig. 18. Change in normalised (dry,15% O 2 ) NO x concentration with water loading

at the lean stability limit.

q

c

l

a

r

v

i

t

a

d

f

a

p

5

i

t

m

t

c

r

o

s

o

e

p

fl

s

l

b

t

c

e

s

c

w

a

l

t

w

a

d

l

w

v

1

o

i

s

t

l

t

i

s

h

e

i

r

e

p

e

o

L

d

s

a

A

i

f

C

u

t

o

t

i

R

uantified for the leanest stable equivalence ratio at each flow

ondition. The results are shown plotted against change in water

oading in Fig. 18 . From initial dry concentrations of ∼12 ppmV

t Ø = 0.6, further normalised reduction by a factor of four is

ecorded for the highest experimental water addition. Again,

apour cases provided marginally lower values as a result of the

nhomogeneous spray mix. CO concentrations were also quan-

ified and demonstrated an equivalent trend of reduction with

n increase in water loading and lowering of equivalence ratio,

ropping from a peak value of ∼18 ppmV at Ø = 0.8 to ∼5 ppmV

or the wettest and leanest conditions. This shift in production is

gain attributed to change in flame temperature and the employed

remixed burner configuration [15,37] .

. Conclusions

As predicted from a recent laminar flame study, subtle changes

n H 2 O fraction had a substantial, counter-intuitive, impact on

he premixed combustion behaviour of a converter gas in a pre-

ixed turbulent, swirling flame. In addition to reducing flame

emperature, H 2 O dissociation enhances chain carrier formation,

atalysing CO oxidation and increasing reactivity and heat release

ate. Changes in vapour H 2 O concentration to levels representative

f fluctuation in atmospheric humidity – up to 1.8% mol – were

hown to enhance laminar flame speed by up to 60%. The impact

f this change on a premixed swirling flame is dependent on

quivalence ratio; near the lean stability limit, an increase com-

resses the area of localised heat release to shorten the elongated

ame structure. However with a stable and well-defined flame

tructure, vapour addition triggers a change in axial heat release

ocation, causing the flame front to retract upstream toward the

urner outlet nozzle.

An atomised spray was employed to deliver equivalent quan-

ities of liquid H 2 O into the CRZ of the premixed flame. Local

atalytic H 2 O dissociation and turbulence resulting from droplet

vaporation combine to provide qualitatively similar but more

ignificant change in heat release. The flame structure is again

ompressed, moving upstream toward the premixed nozzle outlet

ith an increase in H 2 O flowrate. However, compared to vapour

ddition, the spray provides localised narrowing in the heat re-

ease structure. This results from the local influence of H 2 O in

he CRZ, with the flame reverting back to the vapour structure as

ater concentrations are further increased with two-phase flow,

nd the catalytic influence of chain carrying species formation is

iminished. The spray also causes detachment at the lean stability

imit through localised heat loss, compared to the vapour case.

Practical operability of the employed burner configuration

as sensitive to subtle changes in H 2 O loading with either a

apour or liquid spray: Altering overall reactant composition by

.8% mol gave rise to changes in measured surface temperatures

f up to ∼300 K as a function of equivalence ratio. H 2 O induced

ncrease in overall reaction rate also caused a change in the lean

tability limit; excess air could be increased by almost 10% with

he employed configuration at the highest experimental vapour

oading. The observed changes in flame response also highlight

he potential for changes in flame topology or other premixed

nstabilities such as flashback. The permanent addition of a water

pray could dampen fluctuations due to changes in atmospheric

umidity, as the flame was stable with two-phase H 2 O flow. How-

ver, the delivery rate must be carefully specified as the catalytic

nfluence of H 2 O is diminished with excessive supply, and the

eaction quenched.

There is potential for the significant reduction of NO x and CO

missions from using water as a stability mechanism with the em-

loyed flame. NO x concentrations approximately halved for each

xperimental equivalence ratio with loading in either the vapour

r liquid phase. This was further reduced with the reduction in

BO equivalence ratio, lowering flame temperature. Concentrations

ropped to a third of the equivalent dry value for 0.52 g s −1

upply, and were reduced further by a factor of four with the

ddition of both vapour and spray at the highest flow conditions.

cknowledgments

The inspiration for this work was provided by an ongoing

ndustrial partnership with TATA Steel/EU, and supported by

unding from the UK’s Engineering and Physical Sciences Research

ouncil project reference EP/M015300/1 . Information on the data

nderpinning the results presented here, including how to access

hem, can found at Cardiff University data catalogue at http://doi.

rg/10.17035/d.2016.0011507413 . The research was undertaken at

he Cardiff University’s Gas Turbine Research Centre (GTRC) with

nvaluable technical support from Jack Thomas and Terry Treherne.

eferences

[1] R. Remus , M.A. Aguado-Monsonet , S. Roudier , L.D. San , Best available tech-niques reference document on the production of iron and steel, IPPC European

Commission, 2013 Industrial Emissions Directive 2010/75/EU .

[2] D.G. Pugh , Combustion characterisation of compositionally dynamic steelworksgases Ph.D. Thesis, Cardiff University, U.K., 2014 .

[3] A.K. Das , K. Kumar , C. Sung , Laminar flame speeds of moist syngas mixtures,Combust. Flame 158 (2011) 345–353 .

[4] D.G. Pugh , A.P. Crayford , P.J. Bowen , M. Al-Naama , Parametric investigation ofwater loading on heavily carbonaceous syngases, Combust. Flame 164 (2016)

126–136 .

[5] Y. Xie , J. Wang , N. Xu , S. Yu , M. Zhang , Z. Huang , Thermal and chemical effectsof water addition on laminar burning velocity of syngas, Energy Fuel. 28 (2014)

3391–3398 . [6] D.G. Pugh , A.P. Crayford , P.J. Bowen , T. O’Doherty , R. Marsh , Variation in lami-

nar burning velocity and Markstein length with water addition for industriallyproduced syngases, ASME Turbo Expo 2014 (2014) Paper GT2014-25455 .

[7] D. Singh , T. Nishiie , S. Tanvir , L. Qiao , An experimental and kinetic study ofsyngas/air combustion at elevated temperatures and the effect of water addi-

tion, Fuel 94 (2012) 448–456 .

[8] B. Ge , Y. Tian , S. Zang , The effects of humidity on combustion characteristicsof a nonpremixed syngas flame, Int. J. Hydrogen Energy 41 (2016) 9219–9226 .

[9] J. Santner , F. Dryer , Y. Ju , Effect of water content on syngas combustion atelevated pressure, 50th AIAA Aerospace Sciences Meeting (2012) Paper AIAA

2012-0163 .

48 D.G. Pugh et al. / Combustion and Flame 177 (2017) 37–48

[

[

[

[

[

[

[10] A. Bagdanavicius , P.J. Bowen , D. Bradley , M. Lawes , M.S. Mansour , Stretch rateeffects and flame surface densities in premixed turbulent combustion up to

1.25 MPa, Combust. Flame 162 (2015) 4158–4166 . [11] S.A. Filatyev , J.F. Driscoll , C.D. Carter , J.M. Donbar , Measured properties of tur-

bulent premixed flames for model assessment, including burning velocities,stretch rates, and surface densities, Combust. Flame 141 (2005) 1–21 .

[12] C.K Law , Combustion physics, Cambridge University Press, U.K., 2006 . [13] D. Zhao , H. Yamashita , K. Kitagawa , N. Arai , T. Furuhata , Behavior and effect

on NO x formation of OH radical in methane–air diffusion flame with steam

addition, Combust. Flame 130 (2002) 352–360 . [14] G.J. Rørtveit , J.E. Hustad , S. Li , F.A. Williams , Effects of diluents on NO x forma-

tion in hydrogen counterflow flames, Combust. Flame 130 (2002) 48–61 . [15] S. Göke , M. Füri , G. Bourque , B. Bobusch , K. Göckeler , O. Krüger , S. Schimek ,

S. Terhaar , C.O. Paschereit , Influence of steam dilution on the combustion ofnatural gas and hydrogen in premixed and rich-quench-lean combustors, Fuel

Process. Technol. 107 (2013) 14–22 .

[16] G. Cabot , D. Vauchelles , B. Taupin , A. Boukhalfa , Experimental study of leanpremixed turbulent combustion in a scale gas turbine chamber, Exp. Therm.

Fluid Sci. 28 (2004) 683–690 . [17] C.S. Panoutsos , Y. Hardalupas , A.M.K.P. Taylor , Numerical evaluation of equiva-

lence ratio measurement using OH ∗ and CH ∗ chemiluminescence in premixedand non-premixed methane–air flames, Combust. Flame 156 (2009) 273–291 .

[18] M. Lauer , T. Sattelmayer , Heat release calculation in a turbulent swirl flame

from laser and chemiluminescence measurements, 14th International Sympo-sium on Applications of Laser Techniques to Fluid Mechanics (2008) .

[19] N.A. Chigier , J.M. Beer , The flow region near the nozzle in double concentricjets, J. Basic Eng. 4 (1964) 797–804 .

[20] H.J. Sheen , W.J. Chen , S.Y. Jeng , T.L. Huang , Correlation of swirl number for aradial-type swirl generator, Exp. Therm. Fluid Sci. 12 (1996) 4 4 4–451 .

[21] R. Marsh, J. Runyon, A. Giles, S. Morris, D.G. Pugh, A. Valera-Medina, P.J.

Bowen, Premixed methane oxycombustion in nitrogen and carbon dioxide at-mospheres: measurement of operating limits, flame location and emissions,

Proc. Combust. Inst. (2016), doi: 10.1016/j.proci.2016.06.057 . [22] S. Taamallah , S.J. Shanbhogue , A.F. Ghoniem , Turbulent flame stabilization

modes in premixed swirl combustion: physical mechanism and Karlovitz num-ber-based criterion, Combust. Flame 166 (2016) 19–33 .

[23] C. Killer, Abel Inversion Algorithm, http://www.mathworks.com/matlabcentral/

fileexchange/43639- abel- inversion- algorithm , 2013 (accessed 01/12/2015). [24] S.J. Shanbhogue , Y.S. Sanusi , S. Taamallah , M.A. Habib , E.M.A. Mokheimer ,

A.F. Ghoniem , Flame macrostructures, combustion instability and extinctionstrain scaling in swirl-stabilized premixed CH 4 /H 2 combustion, Combust.

Flame 163 (2016) 494–507 .

25] J. Runyon , R. Marsh , Y. Sevcenco , D. Pugh , S. Morris , Development and com-missioning of a chemiluminescence imaging system for an optically-accessible

high-pressure generic swirl burner, 7th European Combustion Meeting (2015)Paper P3-29 .

26] S. Krishna , R. Ravikrishna , Quantitative OH planar laser induced fluorescencediagnostics of syngas and methane combustion in a cavity combustor, Com-

bust. Sci. Technol. 187 (2015) 1661–1682 . [27] U. Stopper , W. Meier , R. Sadanandan , M. Stöhr , G.Bulat M.Aigner , Experimental

study of industrial gas turbine flames including quantification of pressure in-

fluence on flow field, fuel/air premixing and flame shape, Combust. Flame 160(2013) 2103–2118 .

28] British Standard ISO 11042-1:1996, Gas turbines. Exhaust gas emission, Mea-surement and evaluation, British Standards Institution, U.K., 1996 .

29] Delevan Nozzle Technology Report, http://www.delavan.co.uk/pdfs/ntg.pdf , (ac-cessed 01/07/2015).

[30] R.J. Kee , F.M. Rupley , J.A. Miller , M.E. Coltrin , J.F. Grcar , E. Meeks , H.K. Mof-

fat , A.E. Lutz , G. Dixon-Lewis , M.D. Smooke , J. Warnatz , G.H. Evans , R.S. Larson ,R.E. Mitchell , L.R. Petzold , W.C. Reynolds , M. Caracotsios , W.E. Stewart , P. Glar-

borg , C. Wang , O. Adigun , CHEMKIN collection release 3.6, Reaction Design, SanDeigo, USA, 20 0 0 .

[31] CHEMKIN-PRO 15131, Reaction design: San Diego, 2009. http://www.reactiondesign.com/support/help/help _ usage _ and _ support/how-to-cite-

products/ .

32] S.G. Davis , A.V. Joshi , H. Wang , F. Egolfopoulos , An optimized kinetic model ofH 2 /CO combustion, Proc. Combust. Inst. 30 (2005) 1283–1292 .

[33] J. Xia , H. Zhao , A. Megaritis , K.H. Luo , A. Cairns , L.C. Ganippa , Inert-droplet andcombustion effects on turbulence in a diluted diffusion flame, Combust. Flame

160 (2013) 366–383 . [34] T.F. Guiberti , D. Durox , L. Zimmer , T. Schuller , Analysis of topology transitions

of swirl flames interacting with the combustor side wall, Combust. Flame 162

(2015) 4342–4357 . [35] P. Sayad , A. Schönborn , J. Klingmann , Experimental investigation of the stability

limits of premixed syngas-air flames at two moderate swirl numbers, Combust.Flame 164 (2016) 270–282 .

36] T. Furuhata , T. Kawata , N. Mizukoshi , M. Arai , Effect of steam addition path-ways on NO reduction characteristics in a can-type spray combustor, Fuel 89

(2010) 3119–3126 .

[37] H. Kobayashi , S. Yata , Y. Ichikawa , Y. Ogami , Dilution effects of superheatedwater vapor on turbulent premixed flames at high pressure and high temper-

ature, Proc. Combust. Inst. 32 (2009) 2607–2614 .